Protocol

UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026)

Tumour-fraction-stratified cross-validation of fragmentomic feature sets and an ichorCNA tumour-fraction classifier.

Overview

Tumour-fraction-stratified cross-validation of fragmentomic feature sets and an ichorCNA tumour-fraction classifier.

Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.

7 recorded evaluations, 52 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.

View coverage and remaining gaps across all benchmarks

Results

Results are available, but no reviewed comparison panel is linked in this release.

All evaluations

7 evaluations · 52 results. Different protocols are not a single leaderboard.

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.852 accuracy fraction · higher Uncertainty: 95% CI 0.847567448340331 to 0.857161440034147. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 115 (feat 'All', ichorcna_strat 'all', .metric 'Accuracy'), column D 'mean'
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.97 auprc unitless · higher Uncertainty: 95% CI 0.968264347567679 to 0.971522552495379. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 111 (feat 'All', ichorcna_strat 'all', .metric 'AUPRC'), column D 'mean'
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.916 auroc unitless · higher Uncertainty: 95% CI 0.912077380234279 to 0.919242513352284. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 110 (feat 'All', ichorcna_strat 'all', .metric 'AUROC'), column D 'mean'
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.898 f1-score fraction · higher Uncertainty: 95% CI 0.894147470612093 to 0.901005064645811. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 114 (feat 'All', ichorcna_strat 'all', .metric 'F1'), column D 'mean'
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.879 recall fraction · higher Uncertainty: 95% CI 0.872041210173299 to 0.885717327526083. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 112 (feat 'All', ichorcna_strat 'all', .metric 'Sensitivity'), column D 'mean'
Configuration: UNITE XGBoost, All five feature types (UNITE-XGB, model X6) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.779 specificity fraction · higher Uncertainty: 95% CI 0.767944094467296 to 0.791742927566671. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 113 (feat 'All', ichorcna_strat 'all', .metric 'Specificity'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.739 accuracy fraction · higher Uncertainty: 95% CI 0.732531455636329 to 0.745603845619276. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 133 (feat 'CNV', ichorcna_strat 'all', .metric 'Accuracy'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.933 auprc unitless · higher Uncertainty: 95% CI 0.930766088311422 to 0.935304298039049. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 129 (feat 'CNV', ichorcna_strat 'all', .metric 'AUPRC'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.817 auroc unitless · higher Uncertainty: 95% CI 0.811816023264348 to 0.822813746781197. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 128 (feat 'CNV', ichorcna_strat 'all', .metric 'AUROC'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.807 f1-score fraction · higher Uncertainty: 95% CI 0.801436970141717 to 0.81336522207616. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 132 (feat 'CNV', ichorcna_strat 'all', .metric 'F1'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.745 recall fraction · higher Uncertainty: 95% CI 0.732901131538364 to 0.756157326380321. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 130 (feat 'CNV', ichorcna_strat 'all', .metric 'Sensitivity'), column D 'mean'
Configuration: UNITE XGBoost, Copy number aberration (CNA) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.724 specificity fraction · higher Uncertainty: 95% CI 0.699821921055358 to 0.74667535310548. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost Copy number aberration (CNA), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 131 (feat 'CNV', ichorcna_strat 'all', .metric 'Specificity'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.694 accuracy fraction · higher Uncertainty: 95% CI 0.687577912513444 to 0.700234945319011. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 139 (feat 'C/T', ichorcna_strat 'all', .metric 'Accuracy'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.892 auprc unitless · higher Uncertainty: 95% CI 0.889336323734103 to 0.895541264100089. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 135 (feat 'C/T', ichorcna_strat 'all', .metric 'AUPRC'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.744 auroc unitless · higher Uncertainty: 95% CI 0.738865828424514 to 0.748450223017611. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 134 (feat 'C/T', ichorcna_strat 'all', .metric 'AUROC'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.771 f1-score fraction · higher Uncertainty: 95% CI 0.764739366896192 to 0.778687054765388. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 138 (feat 'C/T', ichorcna_strat 'all', .metric 'F1'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.704 recall fraction · higher Uncertainty: 95% CI 0.690319607682921 to 0.71806166813913. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 136 (feat 'C/T', ichorcna_strat 'all', .metric 'Sensitivity'), column D 'mean'
Configuration: UNITE XGBoost, 5' C/T motif ratio per bin (C/T) (Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.667 specificity fraction · higher Uncertainty: 95% CI 0.647183554832278 to 0.684332031420548. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_xgb_x1-x6, row 137 (feat 'C/T', ichorcna_strat 'all', .metric 'Specificity'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.536 accuracy fraction · higher Uncertainty: 95% CI 0.531049658385818 to 0.541013108198761. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 50 (feat 'TF', ichorcna_strat 'all', .metric 'test_acc'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.647 accuracy fraction · higher Uncertainty: 95% CI 0.642731552938907 to 0.650504861551433. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 51 (feat 'TF', ichorcna_strat 'all', .metric 'test_acc_95spe'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.653 accuracy fraction · higher Uncertainty: 95% CI 0.643977997471832 to 0.659762092472169. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 52 (feat 'TF', ichorcna_strat 'all', .metric 'test_acc_98spe'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.634 accuracy fraction · higher Uncertainty: 95% CI 0.616718967325738 to 0.650376938481781. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 53 (feat 'TF', ichorcna_strat 'all', .metric 'test_acc_99spe'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.866 auprc unitless · higher Uncertainty: 95% CI 0.862608963922692 to 0.869544218585427. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 54 (feat 'TF', ichorcna_strat 'all', .metric 'test_auprc'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.642 auroc unitless · higher Uncertainty: 95% CI 0.634830457106185 to 0.649129313044353. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 55 (feat 'TF', ichorcna_strat 'all', .metric 'test_auroc'), column D 'mean'
Configuration: ichorCNA tumour fraction with logistic regression (ichorCNA-TF, Wang et al. 2026)	Protocol: UNITE cross-validation, cancers with all tumour fractions vs healthy (Wang et al. 2026) Dataset: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples	0.552 f1-score fraction · higher Uncertainty: 95% CI 0.544499419390956 to 0.559421831257033. Interval over the 50 outer test folds as printed in columns G and H Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source ichorCNA-TF logistic regression, all tumour fractions ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all Aggregation: Not reported A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing; Wang et al. 2026, Data file S2 (model scores and summary statistics) · Data file S2 sheet STATS_lr, row 56 (feat 'TF', ichorcna_strat 'all', .metric 'test_f1'), column D 'mean'

Source checking is not independent reproduction. Release 2026-10-10-7b8f80935f90.

Methods and evaluation design

Procedure, tasks and evaluated configurations

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Baseline coverage

Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.

0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.

No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.

Author-reported evaluations: 6
External evaluations: 1

Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.

Null control

Proposed control: requires review

Training-set class prior where supervised fitting is permitted

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Conventional reference

Proposed control: requires review

Regularised classifier on simple permitted features, or protocol's conventional reference

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)

Coverage is derived from release 2026-10-10-7b8f80935f90. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.

Run instructions

No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.

Strengths, limitations and unresolved questions

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

0 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-10-7b8f80935f90
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

All evidence (JSONL) (gzip)How to read the evidence

Sources and history

Release 2026-10-10-7b8f80935f90 · Record review: source checked

2 source records and release history

A scalable deep-learning framework for cancer detection using cell-free DNA shallow whole-genome sequencing · Original source · Science Advances 12(28):eady9432, published 2026-07-10; PMC13353424 full-text XML
Wang et al. 2026, Data file S2 (model scores and summary statistics) · Original source · ady9432_data_file_s2.xlsx inside sciadv.ady9432_data_files_s1_and_s2.zip, PMC open-access copy PMC13353424.1

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all

areas: dna-genomes
contexts: clinical_research
protocol: Restrict cancer samples to those with all tumour fractions (healthy controls unchanged), train each model in nested cross-validation (5 outer folds, 10 repeats, StratifiedGroupKFold so no patient is in both training and test folds) and summarise each metric over the 50 outer test folds.
version: Wang et al. 2026 Data file S2 sheets STATS_xgb_x1-x6 and STATS_lr; Methods P47, P49
metric: auroc
metric direction: higher
limitations: Cross-validation within one pooled set of public and new data; not the held-out or unseen test sets.; Tumour fraction strata are defined by ichorCNA, which is also the input of the ichorCNA-TF comparator.; Author-reported by the UNITE developers; the DELFI-XGB comparator is reported only in text and figures.; All strata share the same healthy controls (325 in the cross-validation set); only the cancers change.; The cross-validation set draws 352 samples from the Cristiano et al. 2019 (DELFI) FinaleDB data and 86 from the Jiang et al. FinaleDB liver data, the cohorts of the Hou et al. evaluations and the DELFI judgement.
source locator: Data file S2 rows with ichorcna_strat 'all'; Results P12-P13, P18; Methods P47, P49

Related records

uses data: UNITE cross-validation set: shallow WGS plasma cfDNA, 458 healthy and 1,232 cancer samples
subject: stratification: ctdnafrag-20261009-protocol-wang2026-unite-cv-tf-all
assessment: UNITE XGBoost All five feature types (UNITE-XGB, model X6), all tumour fractions
assessment: UNITE XGBoost Copy number aberration (CNA), all tumour fractions
assessment: UNITE XGBoost 5' C/T motif ratio per bin (C/T), all tumour fractions
assessment: ichorCNA-TF logistic regression, all tumour fractions
assessment: UNITE XGBoost Fragment length (Len), all tumour fractions
assessment: UNITE XGBoost SD of fragment length counts across 5 Mb bins (SD), all tumour fractions
assessment: UNITE XGBoost Short/long fragment length ratio per bin (S/L), all tumour fractions
assessed by: Select a plasma ctDNA fragmentomics detection workflow

Suggest a correction